My name is Sylvester and I’m a student at the University of Toronto working towards my Master’s degree, and my research deals with applying machine learning to medical image processing. I first heard about Zero Gravity Labs (ZGL) when I was five years old and my mother was telling me a story about how it changed her life. Wait sorry maybe she was talking about crockpots in that conversation. The next time was at the Toronto Machine Learning Summit where I ran into Sansom, dressed in his signature red polo sweater, a beacon amongst a storm of summit attendees. He explained that the lab was a space where people had the freedom to explore new ideas, reach outside their comfort zone, and try their hardest to push innovation forward at LoyaltyOne.
What this equated to was a semester of really fantastic conversations, the chance to build machine learning models on production data, review high-impact journal articles, and try new ideas without the fear of failing. As the name implies zero gravity means there’s really nothing to pull you back towards the ground when you jump too high. I guess that could easily translate to meaning untethered or simply being stuck in a motionless state floating into a dark abyss but lets not read too deep into this half-thought-out analogy. What I’m trying to say is that usually there’s someone to give you a slap on the wrist when you don’t get the results you’re looking for, and at ZGL that doesn’t happen. There’s just a group of smart people who ask you how they can help and suggest new ideas to try. The people here are super diverse and there’s a good chance that if you don’t know something or can’t figure it out, somebody has done a multi-year research project on that topic. It’s an intellectual trampoline of sorts, if we are forgoing the zero gravity train of thought.
At ZGL the group is separated into different teams and in this term there was IoT, quantum computing, and the entity embedding team, the latter of which I had the immense privilege to be a part of. We worked on building latent representations of entities at different hierarchies, such as products, baskets, and customers, and then using these embeddings to build ML models for prediction and market segmentation. I had a chance to work with interesting LSTM models, which were a divergence from my previous background in traditional image processing techniques, as well as newer generative techniques.
All in all, it’s been a pretty amazing time. I think the people here are top-notch and the work definitely keeps you on your toes. In that way it’s almost like a ballet dance if you think about it. Maybe that’s why the office dress code involves leotards and tutus. This is a really special place, and I’m looking forward to the future projects that this lab works on and learning about the new and interesting directions they take.
Sylvester Chiang is an a Masters Student at the University of Toronto.
For all Zero Gravity Labs blog content follow us on Medium